Chapter 3 Voice Constructions in Kanakanavu
3.4 Voice and TAM
3.4.3 Mood
Since there are quite a few categories of linguistic mood, we will use conditional and subjunctive as the testing ground. It turns out that most voice constructions are generally judged to be accepted both when conditional and subjunctive moods are present. Below are some examples (a examples are conditional; b examples are subjunctive (with pacupucupuung ‘hope’)).
(72)a. noo te=ku k<um>a-kuun tammi ia if FUT=1SG.NOM Ca<AV>-eat sweet.potato TOP
te=pa=ku k<um>a-kuun ’uru FUT=still=1SG.NOM Ca<AV>-eat rice
‘If I eat sweet potatoes, I will still eat rice.’
b. pacupucupuung=ku k<um>a-kuun tammi nuura think=1SG.NOM Ca<AV>-eat sweet.potato tomorrow
58
‘I hope to eat sweet potatoes tomorrow.’
(73)a. noo vura’u-un=maku tacau ia umori’i if hit-PV=1SG.GEN dog TOP AV.yell ‘If I beat the dog, it’ll yell.’
b. pacupucupuung=ku ia vura’u-un=musu manu=musu think=1SG.NOM TOP hit-PV=2SG.GEN child=2SG.GEN ‘I hope that you hit your child.’
(74)a. noo sia-kuun=maku ’arating iisi ’uru ia urupaca=kasu if IV-eat=1SG.GEN chopsticks this rice TOP use=2SG.NOM tingsi
spoon
‘It I use these chopsticks to eat rice, you use the spoon’
b. pacupucupuung=ku sia-kuun=maku ’arating iisi ’uru think=1SG.NOM IV-eat=1SG.GEN chopsticks this rice ‘I hope to use these chopsticks to eat rice.’
3.5 Summary
In this chapter, we have provided abundant examples of each voice construction, and the relationship between the voice system and the argument structure reveals how speakers introduce different semantic roles in this language. In addition, the compatibility with various TAM markers also tell us about how each voice construction may embed an innate temporal preference and how each interacts with temporal concepts. The following table shows the general results given in Section 3.4.
Table 3.7 Compatibility of Voice and TAM in Kanakanavu
59
Present reading
Past marker ni/cu
Future marker te/tia
Progressive
’esi
Imperfective koo=pa
Conditional/
subjunctive
AV
V V V V V VPV
V V V V V VIV
V V/X V X X VTable 3.7 indicates that the instrumental voice is more restricted and is not as compatible with all the TAM markers as the other two voice types.
In addition to the morphosyntactic description of the voice system, we also want to address the question as to whether the voice reflects any pragmatic implications. In the next chapter, we will attempt to examine how Kanakanavu speakers actually use the voice system on the discourse level, based on the data from our corpus.
60
Chapter 4 Discourse aspects of the voice system
4.1 Introduction
In Chapter 3, the morphosyntax of each voice type in Kanakanavu has been described with examples, but we nevertheless intend to address the question of whether Kanakanavu’s voice system reflects, or is even conditioned by, discourse topicality. In fact, more recent studies show that Austronesian voice system is not simply as a syntactic inflection44, since the system may reveal how discourse aspects play a role in using voice (Wouk 1996, Huang 2002, to name a few). In the present chapter, we hope to find out the relationship between the voice system in Kanakanavu and certain pragmatic factors. More specifically, we will examine the corpus data and see how its voice system may possibly show any correlations to discourse functions such as the individuation of patient argument, grounding, and topicality.
The data analyses in this chapter are established on several pragmatic parameters, e.g., syntactic coding and referential distance, in the case of discourse topicality (cf.
Cooreman et al. 1984). By means of examining eight texts included in NTU Corpus of Formosan Languages (Sung et al. 2008), it is made possible to look at how Kanakanavu’s voice system interacts with pragmatics, if there is indeed a connection.
In fact, we will later discover that Kanakanavu behaves quite differently from highly ergative languages like Tsou, with respect to several discourse properties.
In general, we wish to ask two research questions. First and foremost, does the voice system in Kanakanavu have anything to do with any discourse factors? Several aspects will be investigated in the following sections of this chapter. The results will
44 Some studies have revealed that the Austronesian voice system is a derivation, which means that the voice forms are actually listed in the lexicon. See Reid 1992:67-68, Starosta 1986, Ross 2002:21, etc.
for further discussions.
61
lead to the other question—if we compare with Tsou, a language which is claimed to belong in the same genealogical group with Kanakanavu, would the two languages exhibit a similar patterns in terms of pragmatics? We will reserve the latter important issue for Chapter 5.
The next section gives a sketch of topics which are often associated with Austronesian voice, namely transitivity, ergativity and discourse functions. After that, the methodology and framework adopted in the analyses, along with our findings will be presented in Section 4.3. We will provide certain discourse aspects to which Kanakanavu’s voice system may or may not be sensitive as a whole, and different issues are dealt with in each subsection. Then, we will attempt to put together the results obtained so far and reach an interim conclusion in the last section.
4.2 Austronesian voice, transitivity, ergativity, and discourse
Austronesian voice system is often investigated within various frameworks, and each approach entitles linguists to understand different aspects reflected in the use of voice. We will particularly cover some discussions on transitivity, ergativity, and discourse functions, the last of which will constitute the major topic in this chapter.
One of the fierce debates about Austronesian languages concerns the relationship between voice system and transitivity/ergativity. There are in fact several hypotheses which are assumed by different linguists to be true of the relationship in Philippine-type languages. Some believe that NAV clauses are transitive, while AV clauses are intransitive, hence the ergative hypothesis (Gibson & Starosta 1987, etc.).
Other theories, in contrast to the aforementioned one, claim that NAV clauses are intransitive and AV clauses are transitive; that is, the Philippine-type languages are considered to be syntactically accusative. This claim has been gradually discarded,
62
though. Yet another hypothesis states that both NAV and AV clauses are transitive (Ross 2002:24).
The term ergativity itself is problematic in that it may refer to different notions according to the author. In addition to that, ergativity may be studied on various levels, especially on morphological, syntactic, and discourse levels. Despite the complexities involved, we will select morphological ergativity for the sake of discussion here. To determine whether a language is morphologically ergative is a complex decision.
Consider the following sentences.
(75) t<um>a-tang manu Ca<AV>-cry child
‘The child is crying.’
(76) kuun-un=cu=maku tammi eat-PV=COS=1SG.GEN sweet.potato
‘I already ate the sweet potatoes.’
The sentences above may reveal that, as many Austronesian languages, Kanakanavu may be analyzed as ergative at the morphological level, since the S(ubject) in (75) has the same case alignment with the P(atient) in (76), while the A(gent) in (76) receives a genitive marking. As a matter of fact, AV clauses may involve a semantic patient argument as well, as can be seen in (77).
(77) aririan=ku sua sisiin tamna riang listen.secretly=1SG.NOM OBL bird.type POSS sound
‘I would listen to the sound of the (kind of) bird.’ (from ‘Hunting’)
If these AV clauses are to be interpreted as syntactically intransitive, the patient arguments have to be marked as oblique, much like the Extended Intransitive Clause
63
constructions (EICs).45 Alternatively, the AV clauses with patient arguments may be analyzed as syntactically transitive, and behave more like accusative languages.
Transitivity, on the other hand, may also exert a crucial influence in the use of the voice system as well. The approach proposed by Hopper and Thompson (1980)46 has led to several studies on Philippine-type voice system and case system and comparisons with Indo-European languages. Languages may or may not turn out to be sensitive to the scalar transitivity in terms of the employment of a particular voice type or case marker. As mentioned earlier in the previous chapter, there seem to be some verbs which are expected to be highly semantically intransitive that turn out to be in PV forms. The verb ukusa ‘go’, for instance, may occur with a PV form of a clause, although this PV-marked intransitive verb is required to be followed by another verbal predicate like mara ‘take’, which in effect specifies the primary meaning of the event. See Example (29a) repeated below.
(29)a. ukusu-un=maku m-ara tikuru iisi go-PV=1SG.GEN AV-take clothes this
‘I will go bring the clothes back.’
Sentences like (29a) might be considered at a first glance an exception to our understanding that PV-marked verbs are usually transitive ones. However, on the discourse level, such a misfit can be explained away by the fact that the doer of the action usually initiates the event somewhat due to an outer force. That is, the agent performs the action because of some reason that drives his/her motivation to do it.
45 The concept of EICs is extensively surveyed in other Formosan languages as well. See Huang and Tanangkingsing (2011) for reference.
46 Hopper and Thompson (1980) state that the following 10 parameters show a correlation to transitivity: participants, kinesis, aspect, punctuality, volitionality, affirmation, mode, agency, affectedness of O, and individuation of O. Although the present study will not deal with all of the components, we will investigate the aspect and individuation of O in subsequent sections of this chapter.
64
Hence, example (29a), again, denotes a scenario where the performer (=maku ‘I’) carries out the act of taking back the clothes since, for instance, someone else expects him/her to do so. However, the term transitivity is also notorious for its different uses, and we thus need to clarify on which level, be it semantic or syntactic, the author similar tendency. The following table shows the distribution of AV and NAV forms in Kanakanavu.
Table 4.1 Distribution of voice forms in Kanakanavu
1-argument 2-argument Number Total
AV 185 94 279 (88.0%)
317
NAV 0 38 38 (12.0%)
The table above clearly indicates a tendency in which AV verbs are predominantly in use, whereas NAV ones are a lot lower in frequency and they typically assume a transitive feature since they always involve two or more arguments.
Languages that are claimed to be closer to Proto-Austronesian usually show a fairly frequent occurrence of PV clauses in discourse.
47 Only the main verbs in complete and well-formed clauses are considered. For purposes of simplicity, the following verbal elements are not taken into account in our calculation: predicate-like adjectives, secondary predicates in serial verb constructions, relativized verbs, repeated verbs uttered immediately afterwards/forwards due to speech errors, and auxiliary-like verbs (except tavaru’u ‘can, may’).
48 The eight texts are of various genres included in NTU Corpus of Formosan Languages; the stories are ‘Daily Life’, ‘Hunting Taboos’, ‘Family’, ‘Sowing 1’, ‘Life’, and ‘Mikoong’, ‘Sowing 2, and ‘Pear story’, all of which constitute a total of 622 IUs involving 438 clauses.
65
Nevertheless, the issue of whether Kanakanavu should be treated as an ergative or accusative language is beyond the scope of this study, since both analyses may be developed from equally substantial perspectives. Thus, we hope to bring up these important questions but will leave them to the future research.
In addition to transitivity and ergativity, the voice system in some Austronesian languages may also be subject to factors like topicality. Huang (2002), for instance, explores Tsou and Seediq in terms of the use of voice both in narrative and conversational styles. It is found that AV clauses in Tsou indicate a lower transitivity while NAV clauses suggest a higher one. Furthermore, the use of NAV often marks the [+referential, +definite] feature on the nominative case (or grammatical subject), but the use of AV usually signals the [-referential] feature on NOM argument.
Properties like these will be scrutinized later in Section 4.3.
As a matter of fact, the voice system may turn out to be associated with the following factors, all of which are adopted as parameters examined in Huang’s (2002) work.
a. Discourse ergativity b. Aspect
c. Topicality
Subsequently, we will base our analyses on the three aspects in Kanakanavu since these may be made attainable by means of certain straightforward strategies, as will be explained in Sections 4.3-4.5.
4.3 Discourse ergativity
In this section, we turn to a first discussion in an attempt to see whether the voice system in Kanakanavu has anything to do with discourse ergativity. A language is
66
classified as a discourse ergative language if it meets the following three criteria (Hopper 1982, 1986):
a. PV is the most frequent voice type in natural data.
b. PV correlates with high levels of discourse transitivity.
c. PV is a foregrounding indicator.
Therefore, we will inspect the eight texts again, in order to clarify the discourse status of the voice system in Kanakanavu. Each of the three criteria will be demonstrated and elaborated below.
4.3.1 Frequency of each voice type
It turns out that the most frequent voice form in the eight Kanakanavu texts is agent voice, while non-agent voice (PV and IV) occurs at a much lower frequency.
See Table 4.2 below for the token numbers of each voice form.
Table 4.2 Token number and percentage of each voice type49
No. Text AV (%) PV (%) IV50 Total
1 Daily Life 5 (62.5%) 3 (37.5%) 0 8
2 Hunting Taboos 38 (92.7%) 3 (7.3%) 0 41
3 Family 19 (95%) 1 (5%) 0 20
4 Sowing 1 71 (94.7%) 4 (5.3%) 0 75
5 Life 34 (85%) 6 (15%) 0 40
6 Mikoong 23 (95.8%) 1 (4.2%) 0 24
7 Sowing 2 25 (73.5%) 9 (26.5%) 0 34
49 Again, only the main verbs in complete and well-formed clauses are counted.
50 Instrumental voice does appear sporadically in our corpus, but the instances are all relativized and thus are not counted here.
67
8 Pear story 64 (85.3%) 11 (14.7%) 0 75
Total 279 (88.0%) 38 (12.0%) 0 317
The table clearly displays, without exceptions, a strong tendency of AV uses in all of the narrative stories, which were told by six different native speakers. As a result, unlike discourse ergative languages as discussed in the literature, Kanakanavu shows a decisively dominant use of AV forms (88%), with only around 12% of the occurrence of NAV forms in the texts.
It is somewhat surprising to witness such a low frequency of PV in the statistics since one would probably expect PV to occur equally commonly, or at least not as drastically differently as it appears in Table 4.2. The overwhelming number of AV forms, as opposed to PV, suggests that the two voice types behave distinctly in terms of how speakers use them in natural discourse. The result shown above reflects that Kanakanavu does not satisfy the first criterion of a discourse ergative language. Still, it may be necessary to confirm whether a difference holds in terms of the other two criteria, that is, discourse transitivity and grounding.
4.3.2 NAV and individuation of patients
Hopper and Thompson (1980) hypothesize that the discourse transitivity is manifested largely via the individuation of patients. It is then followed that we may resort to certain properties of the patient argument of a clause in order to investigate the potential relationship between voice and how speakers view this particular argument.
To state more explicitly, the properties that we will look at include referentiality and definiteness, both of which specify how a noun phrase is conceived of and are independent of each other. According to Hopper and Thompson (1980), a noun phrase
68
is considered referential if the speaker is able to single out a referent denoted by the phrase; however, the noun is considered non-referential if the speaker does not mean to refer to any particular referent in any possible worlds. On the other hand, definiteness can be expressed by means of a definite article that tags along the noun phrase. The following are three combinatory types51 derived from the two parameters, with English examples.
Type 1: [+referential, +definite], e.g., ‘I hit the dog sitting over there with a stick.’
Type 2: [+referential, -definite], e.g., ‘I hit a dog with a stick.’
Type 3: [-referential, -definite], e.g., ‘I want to marry a girl (to get married).’
Unfortunately, the article system in Kanakanavu is not as sophisticated as in English, since few, if any, definite/indefinite articles are found in this language. As a result, here we may only make a judgment of which type the patient argument belongs to with respect to the occurrence of demonstrative words (e.g., iisi ‘this’ and iisua
‘that’) and perhaps in a more roundabout way, to the translation.
Before investigating the Kanakanavu data from our corpus, we may first look at a set of elicited sentences that yield a referential and definite pattern as expected.
Below are the examples of each voice type.
(78) Agent voice
a. te=ku k<um>a-kuun tammi FUT=1SG.NOM Ca<AV>-eat sweet.potato
‘I will eat sweet potatoes.’
b. k<um>a-kuun=ku tammi Ca<AV>-eat=1SG.NOM sweet.potato
‘I eat sweet potatoes. / I am eating sweet potatoes.’
51 A fourth type, [-referential, +definite], is considered non-existent, and is thus excluded here.
69
(79) Patient voice
kuun-un=maku tammi eat-PV=1SG.GEN sweet.potato
‘I will eat the sweet potatoes. / I am eating the sweet potatoes.’
(80) Instrumental voice
si-kuun=maku tammi ’arating iisi IV-eat=1SG.GEN sweet.potato chopstick this
‘I use these chopsticks to eat sweet potatoes.’
In sentence (78), which is AV-marked, the patient argument is interpreted as having the [-referential, -definite] feature. This sentence indicates that the speaker will perform the action of eating sweet potatoes, without implying any pragmatically salient role of tammi ‘sweet potatoes’. What makes the case even clearer is the sentence (78a), which denotes that the speaker is one who eats sweet potatoes (as opposed to those who do not). Example (78b), in contrast, does not place a pragmatic emphasis on the patient argument, but it rather refers to the event as a whole.
On the contrary, the patient argument of a PV clause, as in sentence (79), often shows the [+referential, +definite] feature. To put it simply, the speaker has some specific referent(s) of the patient tammi on the mind, and thus, he/she is eating perhaps the sweet potatoes at their sight, or the sweet potatoes which are being talked about. To summarize, the patient argument of a PV clause seems to have a higher referentiality and higher definiteness, since the speaker is talking about some specific referent(s) of the noun phrase.
If we look at other NAV clauses, like the one in (80), it becomes evident that the grammatical subject ’arating ‘chopsticks’, namely the argument triggered by the voice affix, is high in referentiality and definiteness. Kanakanavu speakers prefer IV
70
clauses with a demonstrative in the instrumental nominal phrase, and sentence (81) is thus interpreted as more natural with iisi ‘this’ modifying ’arating ‘chopsticks’ than the one without a demonstrative. It arouses our interests because the features of high referentiality and high definiteness are no longer retained on the part of the object patient argument, but instead on the applied patient, a.k.a., the grammatical subject.
The case of IV clauses, therefore, shows a situation where the grammatical subject—the instrument used to carry out the action—is meant to have a specific referent on the speaker’s mind.
Having examined the elicited data, now we may further ask a question: Will a similar pattern prevail in the texts from our corpus? To answer this, we need to turn to the corpus data now and find out a possible connection between voice types and the referentiality/definiteness of the patient argument. Here, the eight texts are chosen for an illustration. See the examples taken from the corpus.
(81) From Daily Life
[...] er u masiin sua umi ia acecu ni-pu'a-a.
FIL FIL now NOM plum TOP already PFV.PV-buy-REL
FIL FIL now NOM plum TOP already PFV.PV-buy-REL